Audio-Text-to-Text
PEFT
Safetensors
English
moss_audio
audio-captioning
vocal-bursts
lora
custom_code
Instructions to use laion/moss-voice-captioner with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use laion/moss-voice-captioner with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
Document measured locator and temperature search with recommended settings, scores and inference controls
07cad84 verified Download evaluation/locator_decoding_pilot.json from laion/moss-voice-captioner: direct link, hf CLI and curl.
- Browser
- Download file 104 kB
-
https://huggingface.co/laion/moss-voice-captioner/resolve/main/evaluation/locator_decoding_pilot.json
- Command line
-
hf download hf://laion/moss-voice-captioner/evaluation/locator_decoding_pilot.json
-
curl -L -o locator_decoding_pilot.json https://huggingface.co/laion/moss-voice-captioner/resolve/main/evaluation/locator_decoding_pilot.json
104 kB
| { | |
| "schema": "moss-locator-decoding-pilot-v1", | |
| "created_date": "2026-10-11", | |
| "automated_audio_judge": true, | |
| "human_ground_truth": false, | |
| "judge_temperature": 0, | |
| "scale": "0 to 5; 5 is fully correct/useful", | |
| "selection_n": 10, | |
| "confirmation_n": 10, | |
| "sources_total": { | |
| "synthetic_acting": 10, | |
| "emilia_multilingual": 10 | |
| }, | |
| "generation_grid": { | |
| "onset_thresholds": [ | |
| 0.5, | |
| 0.6, | |
| 0.7, | |
| 0.8, | |
| 0.9 | |
| ], | |
| "temperatures": [ | |
| 0, | |
| 0.3, | |
| 0.7 | |
| ], | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "minimum_duration_s": 0.08, | |
| "event_count_cap": null, | |
| "sampling_repeats": 1 | |
| }, | |
| "full_audio_for_both_captioners": true, | |
| "whole_caption_prompt_no_interval": true, | |
| "burst_caption_prompt_requires_start_and_end": true, | |
| "no_script_reference_or_type_label_in_moss_input": true, | |
| "weights": { | |
| "voice_caption_fidelity": 0.3, | |
| "burst_caption_fidelity": 0.25, | |
| "timing_accuracy": 0.2, | |
| "event_precision": 0.15, | |
| "event_recall": 0.1 | |
| }, | |
| "empty_burst_caption_policy": "A clip with no proposed captions scores 0 for burst-caption-set fidelity if the judge heard events; 5 if the judge heard no events. Conditional event averages exclude empty clips.", | |
| "selected_on_selection_clips": "t50-d00", | |
| "selection_results": { | |
| "t50-d00": { | |
| "n_clips": 10, | |
| "n_events": 36, | |
| "weighted_score": 3.5679761904761906, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 3.491904761904762, | |
| "timing_accuracy": 3.4, | |
| "event_precision": 3.5, | |
| "event_recall": 3.8, | |
| "burst_fidelity_event_average": 3.361111111111111, | |
| "config": { | |
| "id": "t50-d00", | |
| "threshold": 0.5, | |
| "temperature": 0.0, | |
| "top_p": null, | |
| "top_k": null, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t50-d30": { | |
| "n_clips": 10, | |
| "n_events": 36, | |
| "weighted_score": 3.4546428571428573, | |
| "voice_caption_fidelity": 3.4, | |
| "burst_caption_fidelity": 3.3985714285714286, | |
| "timing_accuracy": 3.4, | |
| "event_precision": 3.5, | |
| "event_recall": 3.8, | |
| "burst_fidelity_event_average": 3.3333333333333335, | |
| "config": { | |
| "id": "t50-d30", | |
| "threshold": 0.5, | |
| "temperature": 0.3, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t50-d70": { | |
| "n_clips": 10, | |
| "n_events": 36, | |
| "weighted_score": 3.559047619047619, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 3.4561904761904763, | |
| "timing_accuracy": 3.4, | |
| "event_precision": 3.5, | |
| "event_recall": 3.8, | |
| "burst_fidelity_event_average": 3.3333333333333335, | |
| "config": { | |
| "id": "t50-d70", | |
| "threshold": 0.5, | |
| "temperature": 0.7, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t60-d00": { | |
| "n_clips": 10, | |
| "n_events": 29, | |
| "weighted_score": 3.2374404761904763, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 2.689761904761905, | |
| "timing_accuracy": 3, | |
| "event_precision": 3.5, | |
| "event_recall": 3.3, | |
| "burst_fidelity_event_average": 3.3448275862068964, | |
| "config": { | |
| "id": "t60-d00", | |
| "threshold": 0.6, | |
| "temperature": 0.0, | |
| "top_p": null, | |
| "top_k": null, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t60-d30": { | |
| "n_clips": 10, | |
| "n_events": 29, | |
| "weighted_score": 3.148482142857143, | |
| "voice_caption_fidelity": 3.4, | |
| "burst_caption_fidelity": 2.6939285714285717, | |
| "timing_accuracy": 3, | |
| "event_precision": 3.5, | |
| "event_recall": 3.3, | |
| "burst_fidelity_event_average": 3.3448275862068964, | |
| "config": { | |
| "id": "t60-d30", | |
| "threshold": 0.6, | |
| "temperature": 0.3, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t60-d70": { | |
| "n_clips": 10, | |
| "n_events": 29, | |
| "weighted_score": 3.2307738095238094, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 2.663095238095238, | |
| "timing_accuracy": 3, | |
| "event_precision": 3.5, | |
| "event_recall": 3.3, | |
| "burst_fidelity_event_average": 3.3448275862068964, | |
| "config": { | |
| "id": "t60-d70", | |
| "threshold": 0.6, | |
| "temperature": 0.7, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t70-d00": { | |
| "n_clips": 10, | |
| "n_events": 17, | |
| "weighted_score": 3.302857142857143, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 2.8314285714285714, | |
| "timing_accuracy": 3.1, | |
| "event_precision": 3.7, | |
| "event_recall": 3.1, | |
| "burst_fidelity_event_average": 3.7058823529411766, | |
| "config": { | |
| "id": "t70-d00", | |
| "threshold": 0.7, | |
| "temperature": 0.0, | |
| "top_p": null, | |
| "top_k": null, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t70-d30": { | |
| "n_clips": 10, | |
| "n_events": 17, | |
| "weighted_score": 3.192857142857143, | |
| "voice_caption_fidelity": 3.4, | |
| "burst_caption_fidelity": 2.7514285714285713, | |
| "timing_accuracy": 3.1, | |
| "event_precision": 3.7, | |
| "event_recall": 3.1, | |
| "burst_fidelity_event_average": 3.7058823529411766, | |
| "config": { | |
| "id": "t70-d30", | |
| "threshold": 0.7, | |
| "temperature": 0.3, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t70-d70": { | |
| "n_clips": 10, | |
| "n_events": 17, | |
| "weighted_score": 3.3111904761904762, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 2.8647619047619046, | |
| "timing_accuracy": 3.1, | |
| "event_precision": 3.7, | |
| "event_recall": 3.1, | |
| "burst_fidelity_event_average": 3.6470588235294117, | |
| "config": { | |
| "id": "t70-d70", | |
| "threshold": 0.7, | |
| "temperature": 0.7, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t80-d00": { | |
| "n_clips": 10, | |
| "n_events": 7, | |
| "weighted_score": 3.3075, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 2.85, | |
| "timing_accuracy": 3.2, | |
| "event_precision": 3.9, | |
| "event_recall": 2.6, | |
| "burst_fidelity_event_average": 3.7142857142857144, | |
| "config": { | |
| "id": "t80-d00", | |
| "threshold": 0.8, | |
| "temperature": 0.0, | |
| "top_p": null, | |
| "top_k": null, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t80-d30": { | |
| "n_clips": 10, | |
| "n_events": 7, | |
| "weighted_score": 3.1925, | |
| "voice_caption_fidelity": 3.4, | |
| "burst_caption_fidelity": 2.75, | |
| "timing_accuracy": 3.2, | |
| "event_precision": 3.9, | |
| "event_recall": 2.6, | |
| "burst_fidelity_event_average": 3.5714285714285716, | |
| "config": { | |
| "id": "t80-d30", | |
| "threshold": 0.8, | |
| "temperature": 0.3, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t80-d70": { | |
| "n_clips": 10, | |
| "n_events": 7, | |
| "weighted_score": 3.3449999999999998, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 3.2, | |
| "event_precision": 3.9, | |
| "event_recall": 2.6, | |
| "burst_fidelity_event_average": 4, | |
| "config": { | |
| "id": "t80-d70", | |
| "threshold": 0.8, | |
| "temperature": 0.7, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t90-d00": { | |
| "n_clips": 10, | |
| "n_events": 1, | |
| "weighted_score": 2.775, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 1.4, | |
| "timing_accuracy": 2.6, | |
| "event_precision": 3.7, | |
| "event_recall": 2.4, | |
| "burst_fidelity_event_average": 4, | |
| "config": { | |
| "id": "t90-d00", | |
| "threshold": 0.9, | |
| "temperature": 0.0, | |
| "top_p": null, | |
| "top_k": null, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t90-d30": { | |
| "n_clips": 10, | |
| "n_events": 1, | |
| "weighted_score": 2.685, | |
| "voice_caption_fidelity": 3.4, | |
| "burst_caption_fidelity": 1.4, | |
| "timing_accuracy": 2.6, | |
| "event_precision": 3.7, | |
| "event_recall": 2.4, | |
| "burst_fidelity_event_average": 4, | |
| "config": { | |
| "id": "t90-d30", | |
| "threshold": 0.9, | |
| "temperature": 0.3, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t90-d70": { | |
| "n_clips": 10, | |
| "n_events": 1, | |
| "weighted_score": 2.775, | |
| "voice_caption_fidelity": 3.7, | |
| "burst_caption_fidelity": 1.4, | |
| "timing_accuracy": 2.6, | |
| "event_precision": 3.7, | |
| "event_recall": 2.4, | |
| "burst_fidelity_event_average": 4, | |
| "config": { | |
| "id": "t90-d70", | |
| "threshold": 0.9, | |
| "temperature": 0.7, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| } | |
| }, | |
| "confirmation_results": { | |
| "t50-d00": { | |
| "n_clips": 10, | |
| "n_events": 74, | |
| "weighted_score": 3.5487752525252523, | |
| "voice_caption_fidelity": 3.9, | |
| "burst_caption_fidelity": 3.43510101010101, | |
| "timing_accuracy": 3.3, | |
| "event_precision": 3.2, | |
| "event_recall": 3.8, | |
| "burst_fidelity_event_average": 3.3783783783783785, | |
| "config": { | |
| "id": "t50-d00", | |
| "threshold": 0.5, | |
| "temperature": 0.0, | |
| "top_p": null, | |
| "top_k": null, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t50-d30": { | |
| "n_clips": 10, | |
| "n_events": 74, | |
| "weighted_score": 3.5074242424242423, | |
| "voice_caption_fidelity": 3.8, | |
| "burst_caption_fidelity": 3.3896969696969697, | |
| "timing_accuracy": 3.3, | |
| "event_precision": 3.2, | |
| "event_recall": 3.8, | |
| "burst_fidelity_event_average": 3.310810810810811, | |
| "config": { | |
| "id": "t50-d30", | |
| "threshold": 0.5, | |
| "temperature": 0.3, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t50-d70": { | |
| "n_clips": 10, | |
| "n_events": 74, | |
| "weighted_score": 3.501729797979798, | |
| "voice_caption_fidelity": 3.8, | |
| "burst_caption_fidelity": 3.366919191919192, | |
| "timing_accuracy": 3.3, | |
| "event_precision": 3.2, | |
| "event_recall": 3.8, | |
| "burst_fidelity_event_average": 3.27027027027027, | |
| "config": { | |
| "id": "t50-d70", | |
| "threshold": 0.5, | |
| "temperature": 0.7, | |
| "top_p": 0.9, | |
| "top_k": 0, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t60-d00": { | |
| "n_clips": 10, | |
| "n_events": 63, | |
| "weighted_score": 3.6024305555555554, | |
| "voice_caption_fidelity": 3.9, | |
| "burst_caption_fidelity": 3.609722222222222, | |
| "timing_accuracy": 3.3, | |
| "event_precision": 3.4, | |
| "event_recall": 3.6, | |
| "burst_fidelity_event_average": 3.5238095238095237, | |
| "config": { | |
| "id": "t60-d00", | |
| "threshold": 0.6, | |
| "temperature": 0.0, | |
| "top_p": null, | |
| "top_k": null, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| }, | |
| "t70-d00": { | |
| "n_clips": 10, | |
| "n_events": 50, | |
| "weighted_score": 3.4058333333333333, | |
| "voice_caption_fidelity": 3.9, | |
| "burst_caption_fidelity": 2.9233333333333333, | |
| "timing_accuracy": 3.2, | |
| "event_precision": 3.7, | |
| "event_recall": 3.1, | |
| "burst_fidelity_event_average": 3.54, | |
| "config": { | |
| "id": "t70-d00", | |
| "threshold": 0.7, | |
| "temperature": 0.0, | |
| "top_p": null, | |
| "top_k": null, | |
| "nms_overlap_fraction": 0.5, | |
| "min_duration_s": 0.08, | |
| "max_events": null | |
| } | |
| } | |
| }, | |
| "combined_descriptive_results": { | |
| "t50-d00": { | |
| "n_clips": 20, | |
| "n_events": 110, | |
| "weighted_score": 3.5583757215007217, | |
| "voice_caption_fidelity": 3.8, | |
| "burst_caption_fidelity": 3.463502886002886, | |
| "timing_accuracy": 3.35, | |
| "event_precision": 3.35, | |
| "event_recall": 3.8, | |
| "burst_fidelity_event_average": 3.3727272727272726 | |
| }, | |
| "t50-d30": { | |
| "n_clips": 20, | |
| "n_events": 110, | |
| "weighted_score": 3.48103354978355, | |
| "voice_caption_fidelity": 3.6, | |
| "burst_caption_fidelity": 3.394134199134199, | |
| "timing_accuracy": 3.35, | |
| "event_precision": 3.35, | |
| "event_recall": 3.8, | |
| "burst_fidelity_event_average": 3.3181818181818183 | |
| }, | |
| "t50-d70": { | |
| "n_clips": 20, | |
| "n_events": 110, | |
| "weighted_score": 3.5303887085137085, | |
| "voice_caption_fidelity": 3.75, | |
| "burst_caption_fidelity": 3.411554834054834, | |
| "timing_accuracy": 3.35, | |
| "event_precision": 3.35, | |
| "event_recall": 3.8, | |
| "burst_fidelity_event_average": 3.290909090909091 | |
| }, | |
| "t60-d00": { | |
| "n_clips": 20, | |
| "n_events": 92, | |
| "weighted_score": 3.419935515873016, | |
| "voice_caption_fidelity": 3.8, | |
| "burst_caption_fidelity": 3.1497420634920634, | |
| "timing_accuracy": 3.15, | |
| "event_precision": 3.45, | |
| "event_recall": 3.45, | |
| "burst_fidelity_event_average": 3.467391304347826 | |
| }, | |
| "t70-d00": { | |
| "n_clips": 20, | |
| "n_events": 67, | |
| "weighted_score": 3.354345238095238, | |
| "voice_caption_fidelity": 3.8, | |
| "burst_caption_fidelity": 2.8773809523809524, | |
| "timing_accuracy": 3.15, | |
| "event_precision": 3.7, | |
| "event_recall": 3.1, | |
| "burst_fidelity_event_average": 3.582089552238806 | |
| } | |
| }, | |
| "paired_score_delta_t50_minus_t60": { | |
| "mean": 0.13844020562770562, | |
| "bootstrap_95pct_interval": [ | |
| -0.0746969696969697, | |
| 0.40927353896103896 | |
| ], | |
| "n": 20, | |
| "descriptive_only": true | |
| }, | |
| "per_clip_results": [ | |
| { | |
| "clip_sha256": "5db905abb076fb5d3145146e40b9a2266472ab05148f7e341521e105404e9cde", | |
| "source": "synthetic_acting", | |
| "language": "en", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.3333333333333335, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.4833333333333334, | |
| "n_events": 6, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.5, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.525, | |
| "n_events": 6, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.3333333333333335, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.4833333333333334, | |
| "n_events": 6, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.3333333333333335, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.4833333333333334, | |
| "n_events": 6, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.5, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.525, | |
| "n_events": 6, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.3333333333333335, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.4833333333333334, | |
| "n_events": 6, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.6, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.55, | |
| "n_events": 5, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.8, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.6, | |
| "n_events": 5, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.6, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.55, | |
| "n_events": 5, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 2, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 2, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 2, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "d64cdc6d81390d9103bfbf6bfacd5bb7738102a4af6f412753ec1b0063d40bf1", | |
| "source": "synthetic_acting", | |
| "language": "en", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.5714285714285716, | |
| "timing_accuracy": 3, | |
| "event_precision": 2, | |
| "event_recall": 4, | |
| "weighted_score": 3.142857142857143, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 2.5714285714285716, | |
| "timing_accuracy": 3, | |
| "event_precision": 2, | |
| "event_recall": 4, | |
| "weighted_score": 2.842857142857143, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 2, | |
| "event_recall": 4, | |
| "weighted_score": 3.178571428571429, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.6, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.3, | |
| "n_events": 5, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 2.6, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.0, | |
| "n_events": 5, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.4, | |
| "n_events": 5, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.55, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.25, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.55, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.25, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 1.95, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "7a9e39f73cde1ab2e1c78a2839a079b763ec2f399f8eb911accd61f485916e76", | |
| "source": "synthetic_acting", | |
| "language": "en", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 4, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 4, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.575, | |
| "n_events": 4, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.9, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.9, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.3333333333333335, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.4333333333333336, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.9, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.9, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.3333333333333335, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.4333333333333336, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.5, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "a3a4211c1cad74cfa2728a5ba09d354356820d0d3a82e079687ec817b8bc772c", | |
| "source": "emilia_multilingual", | |
| "language": "de", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 2, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.65, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "ebde3d917aa4c72f83f796b95d28df7ed2a95bb8c40f6222d929e8116715eaab", | |
| "source": "emilia_multilingual", | |
| "language": "en", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 5, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "d3762094e1746465f70709fa06eb65911aecea400d1a07cae03533bdfccd4052", | |
| "source": "synthetic_acting", | |
| "language": "en", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.3, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.475, | |
| "n_events": 10, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.2, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.15, | |
| "n_events": 10, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.3, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.475, | |
| "n_events": 10, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.25, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.4625, | |
| "n_events": 8, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.125, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.13125, | |
| "n_events": 8, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.25, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.4625, | |
| "n_events": 8, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.25, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.25, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "017b6b9b611a714b45d1aa8fd540e0ff0d9e96be6bad456d406fb904bb93836b", | |
| "source": "emilia_multilingual", | |
| "language": "de", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "c67c47651c35b58265257342229ba8b7742d9aa97567ab789a18309ea58791d0", | |
| "source": "emilia_multilingual", | |
| "language": "de", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 3.45, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 1.95, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 1.95, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 1.95, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 1.95, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "95cebe7c9810b2199042c2f5d59fea98d4685f598731721ca25fac6df1668645", | |
| "source": "synthetic_acting", | |
| "language": "de", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5785714285714287, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5785714285714287, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.2785714285714285, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5785714285714287, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5785714285714287, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.2785714285714285, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5785714285714287, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5785714285714287, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.7142857142857144, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.2785714285714285, | |
| "n_events": 7, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.5, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.475, | |
| "n_events": 2, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.5, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.475, | |
| "n_events": 2, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.3, | |
| "n_events": 2, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 2.25, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 3, | |
| "event_recall": 2, | |
| "weighted_score": 1.95, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "8c93cf019f4cea3d7e33a969aa359473fa93d17fadcbb88f1892ed8707ad6df7", | |
| "source": "emilia_multilingual", | |
| "language": "en", | |
| "partition": "selection", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t80-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t90-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "f656a894d0f1e14202892d91f13ce353e964f3155b5801da20882cb039206d0d", | |
| "source": "emilia_multilingual", | |
| "language": "de", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.6666666666666665, | |
| "timing_accuracy": 3, | |
| "event_precision": 2, | |
| "event_recall": 4, | |
| "weighted_score": 3.1666666666666665, | |
| "n_events": 9, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.4444444444444446, | |
| "timing_accuracy": 3, | |
| "event_precision": 2, | |
| "event_recall": 4, | |
| "weighted_score": 3.111111111111111, | |
| "n_events": 9, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2, | |
| "timing_accuracy": 3, | |
| "event_precision": 2, | |
| "event_recall": 4, | |
| "weighted_score": 3.0, | |
| "n_events": 9, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 3.8, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "c14dca4cb4f8e79cf6a6a86171f5d5f00da8133d05a491be7e329c3d99e9e116", | |
| "source": "synthetic_acting", | |
| "language": "de", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.5555555555555554, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 3, | |
| "weighted_score": 3.188888888888889, | |
| "n_events": 9, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.5555555555555554, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 3, | |
| "weighted_score": 3.188888888888889, | |
| "n_events": 9, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 2.888888888888889, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 3, | |
| "weighted_score": 2.9722222222222223, | |
| "n_events": 9, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.5555555555555554, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 3, | |
| "weighted_score": 3.188888888888889, | |
| "n_events": 9, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.5, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 3, | |
| "weighted_score": 3.175, | |
| "n_events": 8, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "6af576d7f8eaa5931fcb18656eccfb55435405486f273d0c512f9250bb4ac0e9", | |
| "source": "emilia_multilingual", | |
| "language": "fr", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 2.6666666666666665, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.0166666666666666, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 2.6666666666666665, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.0166666666666666, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 2.6666666666666665, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.3166666666666664, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 3, | |
| "weighted_score": 3.0, | |
| "n_events": 2, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.1, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "8ea1e1885527c052de41490e650df7c73a722c66da43b99118af0e588ac818bf", | |
| "source": "emilia_multilingual", | |
| "language": "fr", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.9, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.9, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.9, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.9, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.9, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "83e321e8c018301b0e504361114f56672701ede034e32dec339c513a78ff8856", | |
| "source": "synthetic_acting", | |
| "language": "de", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.4, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5, | |
| "n_events": 10, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.5, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.225, | |
| "n_events": 10, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 3, | |
| "burst_caption_fidelity": 3.5, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.225, | |
| "n_events": 10, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.4, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5, | |
| "n_events": 10, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.4, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5, | |
| "n_events": 10, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "ffd05bde060c23780ef2845c065c5a8e1bd6c76d873b61220e6afe31fa015c77", | |
| "source": "emilia_multilingual", | |
| "language": "en", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 1, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "8c01cf9c5bcb6bc49633f490dbd5909f87e364392bb791b6d11c835692dcc83c", | |
| "source": "emilia_multilingual", | |
| "language": "fr", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 3, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.7, | |
| "n_events": 2, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 0, | |
| "timing_accuracy": 2, | |
| "event_precision": 4, | |
| "event_recall": 2, | |
| "weighted_score": 2.4, | |
| "n_events": 0, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "43226e5a398d212ffb73529450699b3c898441122317f8cf1ab80b0f4c428800", | |
| "source": "synthetic_acting", | |
| "language": "en", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.5454545454545454, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.536363636363636, | |
| "n_events": 11, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.3636363636363638, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.490909090909091, | |
| "n_events": 11, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.3636363636363638, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.490909090909091, | |
| "n_events": 11, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.625, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.55625, | |
| "n_events": 8, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.6666666666666665, | |
| "timing_accuracy": 3, | |
| "event_precision": 4, | |
| "event_recall": 3, | |
| "weighted_score": 3.6166666666666667, | |
| "n_events": 6, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "45d3a7c3f5616d6f3b02149fea9d380b68a3a426fcff7a1d6947fd6b079c8105", | |
| "source": "synthetic_acting", | |
| "language": "de", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.5833333333333335, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5458333333333334, | |
| "n_events": 12, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.5, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.525, | |
| "n_events": 12, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.5833333333333335, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5458333333333334, | |
| "n_events": 12, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.5833333333333335, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5458333333333334, | |
| "n_events": 12, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.6666666666666665, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5666666666666664, | |
| "n_events": 9, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| }, | |
| { | |
| "clip_sha256": "5c952016b61a67f34a005f69adc9c6fba4c5ffb7f69c1b7ee08988ad0684a2e4", | |
| "source": "synthetic_acting", | |
| "language": "de", | |
| "partition": "confirmation", | |
| "config_results": { | |
| "t50-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.933333333333333, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.6333333333333333, | |
| "n_events": 15, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d30": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.8666666666666667, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.6166666666666667, | |
| "n_events": 15, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t50-d70": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.6666666666666665, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.5666666666666664, | |
| "n_events": 15, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t60-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 3.933333333333333, | |
| "timing_accuracy": 3, | |
| "event_precision": 3, | |
| "event_recall": 4, | |
| "weighted_score": 3.6333333333333333, | |
| "n_events": 15, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| }, | |
| "t70-d00": { | |
| "voice_caption_fidelity": 4, | |
| "burst_caption_fidelity": 4, | |
| "timing_accuracy": 4, | |
| "event_precision": 4, | |
| "event_recall": 4, | |
| "weighted_score": 4.0, | |
| "n_events": 14, | |
| "burst_fidelity_empty_policy": "0 if audible events exist; 5 if no audible events; conditional caption-only averages exclude empty clips" | |
| } | |
| } | |
| } | |
| ], | |
| "total_shared_judge_estimated_usd": 0.50333025, | |
| "cost_scope": "All original 50-clip review, procedural searches and MOSS search in the shared ledger; estimated introductory list-price cost, not an invoice.", | |
| "limitations": [ | |
| "Only 20 selected clips, not a representative independent benchmark.", | |
| "The same 20 audio clips were used earlier for procedural tuning; confirmation is held out only from the present MOSS decoding selection.", | |
| "Only one seeded sample per nonzero-temperature configuration.", | |
| "The reviewer is an audio language model, not a human and not manually verified event ground truth.", | |
| "Selection threshold 0.5 and confirmation threshold 0.6 rank differently; no universal optimum established.", | |
| "Scores on the original 50 listening clips are a different set and protocol and are not a paired before/after comparison." | |
| ] | |
| } | |